Table 5.

Performance measures for convolutional neural networks without using any data augmentation. Shown are averages across 10-fold cross-validation and standard deviation of the mean in parentheses.

Classifier	Accuracy [%]	Precision	Recall	F1-Score	AUC
VGG-16	58.7 (2.5)	0.54 (.03)	0.45 (.03)	0.45 (.04)	0.81 (.02)
VGG-19	63.6 (1.6)	0.61 (.02)	0.53 (.03)	0.54 (.03)	0.84 (.01)
ResNet-50	59.6 (2.2)	0.56 (.02)	0.49 (.02)	0.49 (.02)	0.83 (.01)
ResNet-152	59.5 (1.2)	0.54 (.03)	0.47 (.02)	0.48 (.02)	0.83 (.01)
NASNet	64.5 (3.4)	0.62 (.05)	0.52 (.04)	0.54 (.04)	0.85 (.02)
DenseNet-201	65.9 (2.4)	0.65 (.03)	0.55 (.03)	0.57 (.03)	0.86 (.02)
Custom CNN	50.8 (2.4)	0.39 (.04)	0.32 (.01)	0.30 (.02)	0.73 (.01)